Finally, AI your legal team will actually approve.
Runs on your server. Reads your documents. Never phones home. Full stop.
Upload your policy and compliance documents, ask questions in plain language, and get cited answers with full source references — nothing leaves your network.
30 minutes · We query your document type live · No slides
Published Benchmark
100%
47 of 47 questions passed on the published PDPA compliance benchmark. Evaluated using RAGAS — an independent retrieval-augmented generation scoring framework. Benchmark methodology, test questions, and full results are published here.
Deployment
Your choice
Runs on your own server or on TechnoNet's servers in Malaysia — the same system either way. Book a demo and we will show it on your documents.
Malaysian Enterprise Track Record
30+ Yrs
TechnoNet Sdn Bhd has delivered enterprise IT systems across Malaysia and ASEAN since 1989. KnowledgeCore is built on that same foundation.
Why KnowledgeCore
Three things we never compromise on.
Every design decision in KnowledgeCore comes back to three commitments. Not features — commitments.
01 / 03
Your data never leaves your building.
KnowledgeCore runs entirely on your own server. No cloud connection, no third-party data access, no data residency risk. Full compliance with your organisation's data sovereignty requirements — critical for regulated environments in Malaysia and Brunei.
02 / 03
Answers you can verify.
Every answer cites the exact source document and section. Staff can check the original. Answers come exclusively from your documents — the system cannot invent facts it has not read. If the answer is not in your files, KnowledgeCore says so. It does not guess.
03 / 03
Deployment that fits your timeline.
No AI team required. No ongoing model training. No vendor lock-in.
SMEGuided installer. Drop in your documents. First question answered same day.
EnterpriseWe configure, benchmark, and certify accuracy against your documents before handover.
How It Works
Three steps. No AI expertise required.
From installation to your first verified answer — without a data science team, without cloud accounts, without compromise.
01
Load Your Documents
Upload your SOPs, policy manuals, contracts, HR documents, and technical specifications. KnowledgeCore validates every file before it enters the system — nothing ingests without passing the pre-ingest check.
PDFDOCXXLSX
02
Ask in Plain Language
Staff ask questions the way they would ask a colleague. No search syntax, no keyword guessing, no training required. Any staff member can use it on day one.
03
Get a Cited Answer
KnowledgeCore returns the answer with the source document and section cited. If the answer is not in your documents, it says so — it does not guess. Every answer is traceable.
Second Brain
It knows how your documents connect to each other.
Search returns passages. A second brain returns the chain — which statute
imposes an obligation, what penalty attaches, who enforces it, which instrument amends it. Those
chains run across documents, so they appear in no single passage and no search result can return
them.
A traversal across an English and a Malay statute
s.12A (Act 854)applies toData User
Data Userapplies tos.12A (Act 709)
s.12A (Act 709)part ofPersonal Data Protection Act 2010
Every edge carries the document, section and passage it came
from, so any claim can be traced back to its source.
50 free queries to evaluate on your real documents. Monthly plans from RM 299. Top-up packs never expire. Your server, your data, your control.
SME
Starter · Professional
RM 299 – 999 / month + 8% SST
For small and medium organisations ready to put their documents to work. 10,000–33,000 queries/month. Guided installer, same-day setup, no specialist required.
For large organisations and regulated environments. We configure, benchmark, and certify accuracy against your documents — before you sign off on handover.
KnowledgeCore is architected for environments where data sovereignty, auditability, and access control are non-negotiable — not optional add-ons.
Encryption at Rest
All document vectors, relational facts, and query logs stored encrypted on your server. No plaintext corpus data at rest.
Bearer Token Auth
Per-tenant JWT authentication enforced on every API endpoint. Tenants are fully isolated — no cross-tenant data access is architecturally possible.
Full Audit Logs
Every query logged with timestamp, tenant ID, source documents cited, and answer provenance. Meets audit requirements for regulated industries.
PDPA-Aligned
Data never leaves your server. No cloud API calls, no third-party telemetry, no data residency ambiguity. Fully compatible with Malaysia's PDPA and Brunei's PDPO requirements.
Licence Governance
JWT-signed licence enforcement with tier controls, document limits, and grace-period handling. Governs feature access and deployment scope without network dependency.
Air-Gap Ready
Industrial tier supports fully air-gapped deployment — no internet connection required. All models, dependencies, and licence validation operate completely offline.
FAQ
Questions we get before every demo.
Does my data ever leave my server?
No. KnowledgeCore runs entirely on your own hardware. Documents are ingested, vectorised, and queried locally. No cloud API calls are made during normal operation — not for embeddings, not for inference, not for logging. Your data stays where you put it. See full data flow and security evidence →
What file types does KnowledgeCore support?
Supported: PDF (text-based), DOCX, XLSX. Text extraction, chunking, and semantic indexing happen automatically on upload.
Limitations to be aware of: Scanned PDFs (image-only, no embedded text) require OCR — contact us for scanned document volume. PDF tables and embedded charts are extracted as text; complex visual layouts may lose formatting fidelity. Multimodal ingestion (tables and charts as images) is on the roadmap for engineering and manufacturing clients.
Corpus scale: Starter handles up to 10,000 documents; Professional up to 33,000; Enterprise and Industrial are uncapped. Maximum file size per upload: 100 MB.
How accurate are the answers?
KnowledgeCore passed 100% (47 of 47) on our published PDPA compliance benchmark, and reached 98% on a 43-question academic-policy benchmark. The PDPA benchmark has 47 questions covering breach notification, data subject rights, NCII obligations, penalty thresholds, and cross-framework scenarios. Evaluated using RAGAS (Faithfulness 1.000 — zero hallucinations in any answer). Every answer cites the exact source document and section. We run the same benchmark methodology against your documents before Enterprise handover. View full benchmark report →
How is this different from ChatGPT or Microsoft Copilot?
ChatGPT and Copilot send your documents to external servers for processing. KnowledgeCore runs the entire AI stack — embedding model, language model, vector database — on your own server. No data leaves your network. No subscription to a third-party AI provider. No terms-of-service ambiguity about who owns your data.
Do I need an internet connection to run it?
No. Once installed, KnowledgeCore operates fully offline. The Industrial tier supports air-gapped deployments where the server has no internet access at all — licence validation, model inference, and all queries run without any outbound connection.
How long does setup take?
For Starter and Professional tiers, the guided installer walks through nine steps — typically under two hours on a prepared server. For Enterprise and Industrial tiers, TechnoNet handles the full deployment, document ingestion, and accuracy benchmarking. You sign off on the benchmark result before we hand over the system. See full deployment guide and hardware requirements →
Can it handle multiple departments or tenants?
Yes. KnowledgeCore is multi-tenant by design. Each tenant has its own isolated document collection, vector index, and API credentials. Cross-tenant data access is architecturally impossible — not a policy, a hard technical boundary.
What happens when I update a document?
Re-ingest the updated file. KnowledgeCore will process the new version and the knowledge base reflects it immediately. Incremental ingest with file-hash tracking — so only changed documents are reprocessed — is on the Stage 2 roadmap.
Is there a free trial?
Yes. The Trial tier is free, supports up to 50 documents, and gives you full access to the core Q&A capability. No credit card required. Contact us and we will issue a trial licence key and walk you through the installer.
What languages does it support?
KnowledgeCore is currently optimised for English-language documents and benchmarked on English corpora (100% on the published English benchmark; 98% on a 43-question academic-policy benchmark). An initial Bahasa Malaysia evaluation (20 questions, June 2026) passed every question with no hallucinations, but its correctness score (0.585) is well below English (0.960); a full Bahasa Malaysia benchmark is being prepared. If your documents are primarily in Bahasa Malaysia or mixed BM/English, contact us to discuss your corpus and timeline. See language evaluation status →
Since 1989
Malaysian enterprise IT delivery · 30+ years
SSM Registered
TechnoNet Sdn Bhd · 198901005373 · Petaling Jaya, Selangor
ASEAN-Ready
Registered entities in Malaysia and Brunei Darussalam
Ready to put your documents to work?
No commitment. We will show you KnowledgeCore running on a real corpus.